One student's honest walk through Stanford's famous vision course

What it does

This repo holds one developer’s worked solutions to all three main assignments from Stanford’s CS231n Spring 2017 (Convolutional Neural Networks for Visual Recognition). The assignments cover k-NN, SVM, softmax, neural nets, CNNs, RNNs, and GANs. Both PyTorch and TensorFlow versions are included for assignments 2 and 3.

The interesting bit

The author is upfront about why they did it: to get better at Python and deep learning. No false claims of state-of-the-art results, no rewritten course notes — just the actual notebooks, warts and all. That honesty makes it a safer reference than many “perfect” solution sets that hide their mess.

Key highlights

• All three assignments completed (k-NN through GANs)
• Dual implementations: PyTorch and TensorFlow for assignments 2 and 3
• Direct links to the official course materials at cs231n.github.io
• Author responds to issues — rare for a personal coursework repo
• 600+ stars suggest it has served as a sanity-check for other students

Caveats

• Extra credit tasks remain unfinished (explicitly noted by the author)
• This is one person’s solutions, not official or verified correct
• 2017 course version; some APIs and best practices have shifted since

Verdict

Grab this if you’re currently slogging through CS231n and need to compare your approach when stuck. Skip it if you want a standalone tutorial — you’ll need the lectures and notes for context.